Hedge Funds and the Origins of Private Information in Currency Markets
Transcript of Hedge Funds and the Origins of Private Information in Currency Markets
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 1/28
HEDGE FUNDS AND THE ORIGINS OF PRIVATEINFORMATION IN CURRENCY MARKETS
Carol Osler a and Vitaliy Vandrovych b
Abstract
This paper provides evidence that hedge funds are a critical source of privatefundamental information in currency markets. We analyze the most disaggregated database ofcurrency transactions to date, with ten different categories of market participants including sixcategories of end users. Our analysis of the information content of individual trades indicates thatonly one category of end user has information, specifically hedge funds. Orders placed byinstitutional investors, broker-dealers, central banks and government agencies, largecorporations, and middle-market corporations provide little information about upcoming returns.Orders of banks in every size category carry information, consistent with now-standard theory
that banks gather information from observing customer trades. Theory does not indicate whether banks should be better informed than their customers. Our results suggest that banks are betterinformed than their individual customers, possibly because they aggregate information frommany customers.
February, 2009
a Corresponding author: Brandeis International Business School, Mailstop 32, BrandeisUniversity, Waltham, MA 02454. [email protected].
b State Street Global Advisors, Boston, MA
All errors are the responsibility of the authors.
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 2/28
HEDGE FUNDS AND THE ORIGINS OF PRIVATE INFORMATIONIN CURRENCY MARKETS
This paper addresses two key questions in foreign exchange market microstructure: First,
Who brings private information to the market? Second, What is that nature of that information?
To address these questions we evaluate the information content of orders executed by the Royal
Bank of Scotland, currently ranked fifth among foreign exchange dealing banks worldwide
(Euromoney 2007), over a 16-month period in 2001-2002. Following Anand et al. (2005), among
others, we measure information content as the signed return immediately after execution (“price
impact”) over time horizons ranging from five minutes to two weeks. If orders carry exchange-
rate relevant information then after buy orders for a given currency that currency should
appreciate, and after sell orders it should depreciate.
Our data permit us to disaggregate those placing the orders far more finely than previous
studies. We identify ten groups of market participants: six different groups of end users, or
equivalently bank customers and four groups of banks. The finest previous disaggregation of
end-user order flow considered only three groups of customers: leveraged investors, institutional
investors, and corporate customers (Evans and Lyons 2007). The only other study that
disaggregates interdealer order flow, Bjonnes et al. (2009), likewise disaggregates banks into
four groups.
Standard theory indicates that there is asymmetric information between dealers and their
customers, with customers bringing information to the market (Kyle 1985; Glosten and Milgrom
1985). Consistent with this hypothesis, evidence indicates that customer order flow in foreign
exchange markets does contain information (Evans and Lyons 2007). It has not yet been
ascertained, however, which customers bring the information to the market.
2
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 3/28
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 4/28
aggregate economic activity at home or abroad, or financial factors, such as aggregate risk
aversion or wealth. The real-side information that economic activity is strong, for example, could
be communicated by strong corporate demand for foreign currency. The financial information
that aggregate risk aversion is high, by contrast, could be communicated by low currency
demand from financial customers. In either case, the information itself would be acquired
passively, meaning the corporation or financial institution would not intentionally seek out the
information but would instead reflect the information unknowingly. Individual agents, viewing
their own activity in isolation, might not recognize the broader picture. Dealers, who see order
flow from many agents, could still recognize the broader pattern.By distinguishing among end-user types we can test whether the information in customer
order flow is passively acquired. Passively-acquired information about the real economy would
come from the order flow of large and/or middle-market corporations. Our results indicate,
however, that corporate orders do not carry exchange-rate information. Passively-acquired
information about financial factors they would learn it from the order flow of all three types of
financial customers: institutional investors, broker-dealers, and leveraged investors. Our results
indicate, however, that the orders of institutional investors and broker-dealers do not carry
exchange-rate information. In short, these results suggest that the information carried by foreign
exchange customer order flow is not passively acquired.
The information in customer order flow could also be acquired actively, meaning through
the conscious effort of individuals to anticipate the market. The active trading community –
meaning hedge funds, currency trading associations (CTAs), and the like – devotes much time
and effort to generating private information about exchange rates. Among other things, they
focus intensely on upcoming macro statistical releases, for example, Dealers provide their active
4
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 5/28
customers with frequent summaries of upcoming release dates and times, forecasts of key
figures, and extensive discussion of related macroeconomic developments. Entire firms are
devoted to collecting and disseminating market forecasts. These forecasts are generated by
combining existing public information, their currency traders’ own observations of the world,
and the traders’ own interpretive framework. In finding that leveraged investors are the only end
users whose trades carry information, we provide direct evidence that information in currency
markets is actively acquired.
Indirect evidence for the importance of actively-acquired information comes from two
existing studies. Rime et al. (2007) provides evidence that information about upcoming statisticalreleases is embedded in exchange rates. MacDonald and Marsh (1996) show that there is
heterogeneity in the exchange-rate forecasts of professional forecasts, despite the forecasters’
shared reliance on public information. This heterogeneity “translates into economically
meaningful differences in forecast accuracy,” and is heavily influenced by “the idiosyncratic
interpretation of widely available information” (p. 665). Finally, MacDonald and Marsh show
that heterogeneity has a substantial influence on trading volume.
The hypothesis that leveraged investors are better informed than other end-users is
consistent with important institutional features of currency markets today. Traders at firms that
import and export goods and services, for example, are typically not permitted to speculate in
spot and forward markets. As discussed in Carlson, Dahl, and Osler (2008), this policy reflects
rogue trader risk and associated high administrative costs associated with such speculation. Since
most commercial customers do not trade at high frequencies, the benefits from synthesizing
information to create accurate exchange-rate forecasts is unlikely to outweigh those high costs
for all but the most active commercial firms.
5
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 6/28
Even many financial traders are uninterested in basing their trades on exchange-rate
relevant information – or at least that is the view of many market participants. Taylor and
Farstrup (2006), for example, in their survey of currency management, state:
[T]here are key participants in foreign exchange markets … that are not always seeking profit derived from their currency positions …. [I]n this category are internationalequity managers. While some managers factor in currency positions as they go about
picking foreign stocks, most are attempting to add value through stock, sector, andregion bets rather than currency plays (p. 10, italics in original).
Institutional investors can, of course, justify this focus by citing the familiar academic evidence
that exchange rates approximate a random walk and thus returns cannot be predicted. Broker
dealers typically trade as a representative for individuals and small institutional investors, so theyare no more likely to be informed than such investors. Last but not least, central banks certainly
attempt to monitor the market but, since they are not in the business of profiting from their
trades, there may be no reason to expect those trades to be informed.
This paper has two sections and a conclusion. Section I describes our data. Section II
presents our methodology and discusses our results. Section III concludes.
I. D ATA
Our data comprise all price-contingent orders placed at the Royal Bank of Scotland over
the period June 1, 2001 through September 20, 2002 in three currency pairs: euro-dollar, dollar-
yen, and sterling-dollar. The data include information about each order’s placement time, size,
trigger rate, direction (buy or sell), the “desk” (group of traders) at which the order was placed, a
code identifying the type of agent that placed the order, and the order’s status at the end of the
sample period: executed, deleted, or open. We estimate the execution time as the time the
6
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 7/28
exchange rate (measured at one-minute intervals) first comes within 2 pips of the order’s trigger
rate. 1
Price-contingent orders represent a subset of all order flow. A price-contingent order
instructs a dealer to trade a specified amount at the market price once the currency trades at a
specified price level (“trigger price”). These orders come in two types: stop-loss orders and take-
profit orders. A stop-loss order instructs the dealer to buy (sell) currency if the price rises (falls)
to the trigger. A take-profit order instructs the dealer to buy (sell) if the price falls (rises) to the
trigger. These orders appear to have a significant influence on exchange rates, amplifying the
response to news (Savaser 2006) and creating non-linearities in intraday return dynamics (Osler2003, 2005).
Our ten groups of order-placing agents include six groups of end-users and four groups
from the dealing community. The six end-user groups are: leveraged investors, meaning
primarily hedge funds and similar organizations called Commodity Trading Arrangements or
CTAs; institutional investors, meaning mutual funds, pension funds, and other low-leverage
asset-management firms; large “corporate” (non-financial) customers, examples of which would
be General Electric; middle-market corporate customers; broker dealers like Brown Brothers
Harriman, who undertake forex transactions associated with the securities transactions of clients
who typically range from individual retail investors to modest-sized mutual funds. 2
The four dealing-community groups are: Royal Bank itself, including its spot/forward
dealers at various locations around the world and the exotic options desk, located in London;
other major forex dealers like Citibank or Deutschebank, whom we label Global Liquidity
1 We verified the accuracy of this approach by calculating the frequency of order executions at different timehorizons relative to order placement and comparing that frequency with the known frequency in an earlier dataset.The earlier dataset is described in Osler (2003).2 A few broker dealers in the dataset were known to be trading exclusively on behalf of one or two leveragedinvestors. In consequence, their trades are included among “leveraged investors.”
7
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 8/28
Providers; other medium-sized forex dealers like Santander or PNB, whom we label Regional
Liquidity Providers; and smaller forex dealers who provide little liquidity to the interbank market
but instead trade in the interbank market almost entirely to service customers. 3
Table 1 provides basic descriptive statistics. During our sample period Royal Bank
received 36,806 orders in these currencies worth a total of $193.9 billion. The orders can be
roughly evenly divided into three categories: orders placed by end-users, orders placed by other
dealers, and internal RBS orders.
Table 2 compares placement patterns between stop-loss and take-profit orders, and
between different order sizes. Take-profit orders are more frequent, accounting for 60 percent ofall orders. The mean order size, of about $5 million, does not vary much between stop-loss and
take-profit orders. Since the size distribution is positively skewed, the median order size of $3
million is quite a bit lower than the average order size. The largest placed order is worth $750
million, though only 2 percent of orders were worth $25 million or more. Very few orders (only
ten) were worth less than $100,000. Only 27 percent of orders are actually executed, a figure that
is fairly consistent across order types and order sizes. The largest executed order, worth $473
million, was placed by corporate customer in the euro-dollar market. The average order is open
about 4 days but the median time open is only about 5 hours; the difference indicates that a
minority of orders are open for months.
Table 2 also compares orders placed by customer type. The share of take-profit orders is
highest among corporate customers, reaching 75 percent for large corporations and 83 percent
3 The exotic options dealers place orders in part to hedge their barrier options. These are options that either comeinto existence, or disappear when the underlying exchange rate touches a certain pre-specified level. When suchlevels are crossed dealers are required to make large changes in their forward market delta-hedges. If the optiondisappears, any existing delta-hedge must be unwound; if the option appears, a delta-hedge must be put in place. Theoption dealers can ensure these transactions take place in a timely manner by placing orders with their spot dealers.
8
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 9/28
for middle-market corporations. Global liquidity providers – the very biggest banks – and
broker-dealers are the only groups that execute more stop-loss orders than take profit orders.
Figure 1, panels A-C, graphically outline intraday patterns in the placement and
execution of these orders. Order placement peaks first during Asian trading, then again during
the London morning, and finally a third time during the morning in New York/late afternoon in
London. 4 This pattern reflects, at least in part, the basic biological rhythms associated sleeping
and eating. Order placement falls whenever traders leave work for the day and rises whenever a
market opens and traders return to their desks; similarly, it falls when traders leave for lunch and
rises when they come back.The small number of orders placed during Asian trading reflects the bank’s
London/European base. The high peak in the London afternoon partly reflects the coincidence of
active trading in New York and London. It also, however, reflects the formal placement of orders
previously received by dealers at RBS and other banks as they leave for the day. Many of the
orders would have been received earlier but noted only on a pad of paper; when someone else
needs to monitor the order it must be formally entered into the bank’s computer system.
Order execution has a distinctly different pattern than order placement. Instead of
multiple sharp peaks, it rises to a plateau after London trading begins and only declines again at
the end of London trading. This can be explained by analyzing daily patterns in the two factors
that jointly dominate order execution: (i) the size of the order book; and (ii) exchange-rate
volatility. The order book will be largest from mid-day London through the end of trading for the
day, a period that coincides with the highest exchange-rate volatility, as shown in Figure 2.
The statistical analysis below necessarily focuses on executed orders. Table 3 presents
the descriptive information for number, average size, and total value of executed orders for each
4 Placement patterns for stop-loss and take-profit orders, not shown separately, are strongly similar to each other.
9
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 10/28
customer type. The 9,950 executed orders have average value of roughly $5 millions; in
aggregate these orders were worth roughly $47.5 billion.
II. M ETHODOLOGY , RESULTS , DISCUSSION
This section presents our central tests for the information content of price-contingent
orders. We find that leveraged investors and large corporations are the only end-users whose
orders have a strong price impact. Members of the dealing community appear to have
information similar to that of the leveraged investors. We also find that the price impact of orders
is concave.
A. Methodology
We evaluate the price impact of executed orders, disaggregated by customer- and order-
type (stop-loss, take-profit). Following the literature (e.g., Anand et al. 2005), price impact is
defined as the (log) change in price relative to the order’s trigger rate immediately following the
order’s execution. We consider eight time horizons: five minutes, thirty minutes, one, six, and
twelve hours, one day, one week, and two weeks.
Our baseline regression aggregates executed orders across all three currencies and
assumes that order flow has a linear price impact on returns:
(1) t i
ii
it k t V ss ε δ β ++=− ∑∑==
+TP ti,
SL ti, V **10000*)(
8
1
8
1
The dependent variable is the signed return measured in basis points, since s t is the log of the
exchange rate expressed as foreign currency per USD. Time horizon is indicated by k .
( ) is the signed value of executed stop-loss orders (take-profit orders) of customer type i,
measured in millions of USD. The raw return is multiplied by plus one if the customer ‘buys’ US
SLti,V
TPti,V
10
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 11/28
dollars and by minus one otherwise: if customer purchases (sales) bring higher (lower) prices, as
indicated in the broader literature, the coefficients should have a positive sign.
The assumption that order flow has a linear impact on price is standard in the empirical
literature, but only because it serves as a useful first approximation. There are good reasons to
expect the relationship to be concave, instead. It is widely appreciated that splitting large orders
into smaller individual transactions and timing the execution of each trade carefully can reduce
the impact of large trades. This is demonstrated in the theoretical treatment of Bertsimas and Lo
(1998) and it is standard practice among dealers. Indirect evidence of a concave effect is
provided in Berger et al . (2006), which shows that the proportionate price impact of minute-by-minute interdealer order flow, while positive in all cases, declines with the amount. Hasbrouck
(1991) finds that equity order flow declining proportionate impact.
To examine the possibility of a concave relation between order flow and returns we
postulate a logarithmic relation, which provides a parsimonious alternative to the linear function
form. To conserve the sign of our orders, we take the log of absolute order size and then reassign
the buy-sell direction, and denote the resulting variable log (V i,t x), x ∈ (SL,TP ):5
(2) t i
ii
it k t ss ε δ β ++=− ∑∑==
+ )log(*)log(*10000*)(8
1
8
1
TP ti,
SL ti, V V
B. Results
The results from estimating the linear relation between order flow and returns, Equation
(1), are presented in Table 4; the results from the non-linear relation, Equation (2), are presented
in Table 5. The top of the table shows price impacts for stop-loss orders; the bottom presents
5 We also estimate a quadratic form of this regression, in which we added the signed squared value of each ordersize to the right-hand-side of equation (1). The results, suppressed to save space, indicated a concave price impactfunction but they have the implausible implication that very large orders would have a negative price impact.
11
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 12/28
price impacts for take-profit orders. For each order type, end-user groups are reported first and
dealing-community groups last.
We draw three main conclusions from this analysis. First, leveraged investors are the
only informed end-user type, while all bank categories are informed. Second, banks have more
information than their informed customers. Third, the relation between order flow and returns is
concave.
Who has information? Leveraged investors are the only end users whose orders have a
statistically significant price impact. A $10 million leveraged-investor stop-loss buy order
appreciates the currency by 4.3 basis points after thirty minutes and by 9.9 basis points after sixhours, according to the non-linear estimates. Thereafter the effect is lower and is no longer
statistically significant. There is a strong, positive relation between post-trade returns and
executed stop-loss orders for all four types of dealers. This evidence is consistent with the
hypothesis that leveraged investors are the only end users that bring information to the currency
market, and that dealers benefit from that information.
It is not clear, at a theoretical level, whether banks would be more or less informed than
their customers. Banks necessarily gain an imperfect signal of any given customer’s information,
but by aggregating these noisy signals banks might ultimately achieve a signal that is stronger
than the ones originally perceived by the customers. Our evidence suggests that the aggregation
effect dominates: dealers know more than their individual customers. At any given time horizon,
the executed bank orders generally about the same price impact than leveraged-investor orders,
with the exception of Regional liquidity providers, whose price impact is statistically larger. For
banks, however, the price impact of stop-loss orders generally continues to rise – and to remain
statistically significant – even after the twelve-hour horizon at which leveraged-investor trades
12
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 13/28
are no longer informative (Figure 4A). The bank price impacts remain economically large, and
are generally statistically significant, even at the two-week horizon.
Take-profit orders do not appear to carry information about upcoming returns, whether
they are placed by end-users or by banks. With respect to the end users, for example, only one of
the 96 total coefficients (across the linear and concave regressions) is positive and statistically
significant, as required for the trades to carry information. A few of these coefficients are
negative and significant, of which most concern central banks. This is difficult to interpret,
though it certainly suggests that central banks are not exploiting the market for their own profit.
The results for dealers are similar. Almost all the price-impact coefficients for take-profit ordersare insignificant, and those that reach standard significance levels are generally negative rather
than positive.
Why might stop-loss orders predict returns for informed agents while take-profit orders
do not? We hazard two guesses based on institutional information. First, the use of take-profit
orders may be driven by option-related logic rather than information. Suppose a corporate
customer needs currency to import some intermediate product; alternatively, suppose an
internationally-invested index fund just received an infusion of new funds and, in order to
expand its ownership of the index of some country, it needs to purchase currency. In both cases it
is not necessarily wise to trade at the first price of the day since they could, instead, place a take-
profit order early in the day which gives them a good chance of trading at a more attractive price
later on. (If the order is not executed they would fill the order at the market price at the end of the
day.) The option to trade later is free even though intraday volatility makes it valuable: even with
no information whatsoever about the exchange rate’s likely movements they would be wise to
place the take-profit and exploit that option.
13
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 14/28
Difference in hedging practices across the two order types could also influence measured
post-trade returns. Dealers hedge take-profit orders by placing limit orders in the interdealer
market, a hedge that is so secure that price risk on take-profit orders is conventionally assigned
to the dealers rather than their customers. A limit order for amount Q at price P tells the market
that the dealer is willing to trade up to Q so long as he gets price P or better. Thus limit orders
act as absorbing barriers: A price rise that triggers a customer take-profit sell order at price P
(requiring the dealer to buy the currency in question) would simultaneously trigger the associated
limit sell order at price P placed by the dealer in the interdealer market. This limit sell order
would normally be executed at the limit price P exactly, causing the price to stop moving but notto reverse course. This could explain why take-profit orders have so little price impact in general,
and why any impact they do have is delayed.
Dealers have no useful way to hedge stop-loss orders. Stop-loss orders are not allowed on
the interdealer brokers, and placing a stop-loss order with another dealer would not provide good
protection since price risk on stop-loss orders is, by market convention, borne by the agent
placing the order. Thus when a customer’s stop-loss buy order is triggered by a price rise, the
dealer (who has to sell to the customer) will quickly eliminate the associated short position by
making a purchase in the interdealer market. This generally requires the dealer to choose
between placing a market order or a limit order. A market order seems most likely, since stop-
loss orders can trigger price cascades (Osler 2005) and in general dealers consider these orders
risky (“every stop-loss order is a potential relationship breaker,” says one dealer). So a price rise
that triggers a customer’s stop-loss buy order would thus trigger market buy orders in the
interdealer market, pushing prices up further. 6
6 We note in passing that the two market conventions discussed here are confirmed in separate orders data from theRoyal Bank of Scotland, described in Osler (2003, 2005), that include execution prices.
14
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 15/28
Linear or Concave? A comparison of Tables 4 and 5 shows that our qualitative
conclusions about information are consistent regardless of whether we assume a linear or
concave relation between order flow and returns. Stop-loss orders placed by leveraged investors
and by anyone in the dealing community have a significant impact that begins almost
immediately and broadly rises with time horizon, while take-profit orders do not carry
information.
Though the F statistics are highly statistically significant, the regressions’ explanatory
power, reported at the bottom of each table, is small. This is to be expected, given the high level
of noise commonly associated with financial markets. Explanatory power is almost invariablylow when financial returns are examined at such high frequencies – the most notable exception,
the daily regressions using aggregate interdealer order flow (e.g., Evans and Lyons 2002, Berger
et al. 2006) prove the rule. Their order flow aggregates millions of trades while our regressions
examine single transactions.
The explanatory power varies with time horizon, rising between the 5-minute and 1-hour
horizons, leveling off, and then generally falling. A similar pattern is observed in Berger et al.’s
(2006) analysis of aggregate interdealer order flow.
Notably, the explanatory power is consistently higher when the relation is concave than
when the relation is linear, with the difference averaging roughly ten percent of the linear value.
Figure 4B juxtaposes the estimated price impact of leveraged-investor stop-loss orders from
Equations (2) with corresponding estimates from Equation (1). The price impact values are
roughly the same for orders below $20 million but diverge sharply for larger orders. While the
linear relation predicts that a $100 million buy leveraged-investor stop-loss order brings an 85
basis point appreciation, the non-linear relation predicts an appreciation of only 21 basis points.
15
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 16/28
III. C ONCLUSIONS
This paper provides evidence that leveraged investors are the only type of currency end
user that brings private information to the markets. Our data comprise price-contingent orders
placed at the Royal Bank of Scotland over 16 months during 2001-2002. Our data allow the
agents placing orders to be disaggregated into ten categories: six categories of end-users –
leveraged investors, institutional investors, broker-dealers, large corporations, middle-market
corporations, and governments and central banks – and four categories of banks – Royal Bank
itself, other major banks, regional banks, and smaller banks that primarily service local
customers. We evaluate the price impact of orders placed by each group over time horizons
ranging from five minutes to two weeks. We address three questions: Who brings private
information to the market? Is the information acquired actively or passively? Do dealers know
more or less than their informed customers?
We find that leveraged investors are the only end users whose trades consistently carry
information. Their executed stop-loss orders have statistically and economically significant price
impact, which rises from approximately four basis points per $10 million order at thirty minutes
to ten or more basis points at the one-day horizon. Though it is standard in the literature to
estimate price impact as a linear function of order flow, our results indicate that the price impact
of individual trades is in fact a concave function of order size.
Theory suggests that members of the dealing community learn exchange-rate relevant
information from observing customers order flow (Evans and Lyons 2002). Our results are
consistent with this hypothesis, since stop-loss orders from all four bank groups have a
statistically significant price impact that is statistically indistinguishable from the impact of
leveraged investors’ stop-loss orders at horizons below 12 hours. At longer horizons, however,
16
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 17/28
dealers trades appear to carry information while the leveraged-investor trades do not. This
suggests that dealers, by observing the trades of many customers, are ultimately better informed
than their customers taken individually.
Our finding that corporations and institutional investors do not appear to bring
information to the market suggests that private information is not a passive reflection of existing
economic conditions. Our finding that leveraged investors do bring information to the market
suggests that private information is instead actively acquired through the informed interpretation
of publicly available information.
REFERENCES
Anand Amber, Sugato Chakravarty, Terence Martell. 2005. Empirical Evidence on the Evolutionof Liquidity: Choice of Market Versus Limit Orders by Informed and Uninformed Traders.Journal of Financial Markets 8: 289-309.
Anand, Amber, and Avanidhar Subrahmanyam, 2006. “Information and the Intermediary: AreMarket Intermediaries Informed Traders in Electronic Markets?” Typescript, SyracuseUniversity.
Berger, David, Alain Chaboud, Sergei Chernenko, Edward Howorka, and Jonathan Wright
(2006). “Order Flow and Exchange Rate Dynamics in Electronic Brokerage System Data,”Board of Governors of the Federal Reserve System, International Finance Discussion Papers No. 830.
Bertsimas, Dimitris, and Andrew W. Lo, 1998. Optimal control of execution costs. Journal ofFinancial Markets 1: 1-50.
Carlson, John A., Christian Dahl, and Carol Osler. 2008. Short-run exchange-rate dynamics:Theory and Evidence. Working paper.
Dominguez Kathryn and Freyan Panthaki. 2006. What Defined ‘News’ in Foreign ExchangeMarkets, Journal of International Money and Finance 25: 168-198.
Evans and Lyons. 2002. Order Flow and Exchange Rate Dynamics. Journal of Political Economy 110: 170-180.
Evans and Lyons. 2007. Exchange Rate Fundamentals and Order Flow. NBER Working Paper13151. National Bureau of Economic Research.
Foucault, Thierry, 1999. Order Flow Composition and Trading Costs in a Dynamic Limit OrderMarket, Journal of Financial Markets 2: 99-134.
17
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 18/28
18
Froot Keneth A. and Tarun Ramadorai. 2005. Currency Returns, Intrinsic Value, andInstitutional-Investor Flow. Journal of Finance 60(3): 1535-1566.
Glosten, Lawrence R. and Paul R. Milgrom. 1985. Bid, Ask and Transaction Prices in aSpecialist Market with Heterogeneously Informed Agents. Journal of Financial Economics14(1): 71-100.
Harris, Lawrence, 1998. Optimal dynamic order submission strategies in some stylized trading problems, Financial Markets, Institutions, and Instruments 7: 1-75.
Hasbrouck Joel. 1991. Measuring the information content of stock trades. The Journal ofFinance 46: 179-207.
Ho T. and H.Stoll. 1983. The Dynamics of Dealer Markets Under Competition. Journal ofFinance 38: 1053-1074.
Kyle A. 1985. Continuous Auctions and Insider Trading. Econometrica 53: 1315-1335.
Lyons Richard K. 2001.The Microstructure Approach to Exchange Rates. MIT Press,Cambridge, MA. 333pp.
MacDonald, Ronald and Ian Marsh. 1996. Currency Forecasters are Heterogeneous:Confirmation and Consequences. Journal of International Money and Finance 15(5): 665-685.
Osler Carol. 2003. Currency Orders and Exchange-Rate Dynamics: Explaining the Success ofTechnical Analysis. Journal of Finance. 58: 1791-1819.
______________ 2005. “Stop-Loss Orders and Price Cascades in Currency Markets,’ Journal ofInternational Money and Finance 24: 219-41.
Osler, Carol, 2007. “Foreign Exchange Microstructure: A Survey,” forthcoming in BruceMizrach, Ed., Encyclopedia of Complexity and System Science (New York: Springer
Science & Business Media)Osler, Carol, Alexander Mende, and Lukas Menkhoff, 2007. “Price Discovery in Currency
Markets,” Typescript, Brandeis University
Rime, Dagfinn, Lucio Sarno, and Elvira Sojli, 2007. “Exchange-rate Forecasting, Order Flow,and Macro Information,” Norges Bank Working Paper 2007-2.
Savaser Tanseli. 2006. “Exchange-Rate Response to Macro News: Through the Lens ofMicrostructure,” Typescript, Williams College.
Shleifer, Andrei, 1986. “Do Demand Curves Slope Down?” The Journal of Finance 41: 579-90.
Taylor, Andrew, and Adam Fartstrup (2006), Active Currency Management: Arguments,
Considerations, and Performance for Institutional Investors (CRA RogersCasy InternationalEquity Research: Darien, Connecticut).
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 19/28
Table 1. Placed Orders By Client Type
Data underlying this table comprise all stop-loss and take-profit orders placed at the Royal Bank of Scotland ov1st 2001 till September 20th 2002 in three currency pairs: euro-dollar, sterling-dollar, and dollar-yen. Totals mrounding.
Number oforders
Percent oforders
DollarValue
($ Billions)
Percentof Order
Value
All Orders 36,781 100.0 193.7 100.0End Users 12,061 32.9 62.6 32.3
Leveraged Investors 2,356 6.4 12.6 6.5Institutional Investors 1,352 3.7 7.8 4.0Brokers/Dealers 2,384 6.5 13.7 7.1Central Banks and
Government Agencies 544 1.5 3.9 2.0Large Corporations 3,623 9.9 20.1 10.4
Middle Market Corp. 1,802 4.9 4.5 2.3
Banks 24,720 67.2 104.6 67.7Royal Bank Internal 13,105 35.6 77.3 39.9Global Liquidity Providers 4,162 11.3 19.8 10.2Regional Liquidity Providers 1,352 3.7 7.5 3.9Local Customer-Service
Banks 6,101 16.6 26,5 13.7
19
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 20/28
Table 2. Descriptive Statistics, Placed OrdersData underlying this table comprise all stop-loss and take-profit orders placed at the Royal Bank of Scotland ov1st 2001 till September 20th 2002 in three currency pairs: euro-dollar, sterling-dollar, and dollar-yen. Totals mayrounding.
AllOrders
Stop-Loss
Take-Profit
Very Small Orders(<=$100,000)
Number of Orders 36,781 16,064 20,717 10 Share of Orders (%) 100 43.7 56.3 0.03
Size ($ Mill.): Mean 5.3 5.5 5.1 0.05 Median 3.0 3.1 2.6 0.05
Distance to Mkt. (in pips): Mean 77.4 78.4 76.7 76.1 Median 43.9 40.0 47.0 71.4
Share Within 1/2 std of 1-Day Return (%) 28.7 30.4 27.4 20.00 Share Within 1 std of 1-Day Return (%) 57.5 60.6 55.0 30.00 Share Within 2 std of 1-Day Return (%) 82.9 83.5 82.4 70.00
Days Open: Mean 4.2 3.4 4.7 0.3 Median 0.2 0.1 0.2 0.3
Share Executed (%) 27.1 24.6 29.0 20.0
Share by End-User Category (%):Leveraged Investors 6.4 35.9 64.1 0.00Institutional Investors 3.7 39.8 60.2 0.00Brokers/Dealers 6.5 67.5 33.5 0.12Central Banks and Government Agencies 1.5 27.4 72.6 0.00 Large Corporations 9.9 28.6 71.4 0.00
Middle-Market Corporations 4.9 18.6 81.4 0.03
Share by Bank Category (%):Royal Bank Internal Orders 35.6 44.7 55.3 0.04Global Liquidity Providers 11.3 64.4 35.6 0.00Regional Liquidity Providers 3.7 45.2 54.8 0.00Customer-Service Banks 16.6 37.8 62.2 0.04
20
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 21/28
21
Table 3. Descriptive Statistics, Executed Orders
Data underlying this table comprise all stop-loss and take-profit orders executed by the Royal Bank of Scotland 1st 2001 till September 20th 2002 in three currency pairs: euro-dollar, sterling-dollar, and dollar-yen. Totals mrounding.
Stop-Loss Take-PTotalNumber
Stop-Loss
ShareNumber Average
($ mill.)Total
($ mill.) Number Aver($ m
End-UsersLeveraged Investors 750 28.1 211 5.5 1,168 539 5.2 Institutional Investors 413 41.4 171 4.6 786 242 3.9 Broker/Dealers 485 59.4 288 4.6 1,333 197 4.7 Central Banks and
Government Agencies 164 31.1 51 3.9 199 113 7.9
Large Corporations 1,274 24.7 315 6.6 2,089 959 4.4
Middle-Market Corporations 754 17.5 132 2.9 291 622 1.9
BanksRoyal Bank Internal 3,527 38.8 1,368 6.2 8,501 2,159 5.2Global Liquidity Providers 942 68.9 649 4.0 2,604 293 4.0Regional Liquidity Providers 315 43.5 137 5.8 795 178 5.3Customer-Service
Providers 1,325 47.4 628 4.6 2,888 697 3.4
Total 9.859 40.1 3,950 5.3 20,747 5,909 4.
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 22/28
Table 4. Price Impact of Price-Contingent Orders: Linear Estimates This table summarizes the results of the regressions testing for a difference in the price impact of theorders from different client types:
t i
ii
it k t V ss ε δ β ++=− ∑∑==
+TP ti,
SL ti, V **10000*)(
8
1
8
1
,
where st is the log exchange rate expressed as foreign currency units per USD. k is the time horizon overwhich we measure price impact. is the signed dollar value of stop-loss order of customer of the type
i; is the signed dollar value of take-profit order of customer of the type i. Size is positive (negative)
if the client instructs the dealer to buy the commodity (denominator) currency. Coefficients can beinterpreted as the impact of $1 million in basis points. * 10 percent significance; ** 5 percentsignificance; *** 1 percent significance.
SLti,V
TPti,V
5 min 30 min 1 hour 6 hours 12 hours 1 day 1 week 2 weeksStop-Loss Orders
End-UsersLeveraged Investors 0.058 0.173** 0.163* 0.486*** 0.269 0.229 0.166 0.292
Institutional Investors -0.055 -0.236 -0.392 -0.249 -0.533** -0.275 -0.219 -0.739Broker-Dealers -0.043 0.096 0.032 0.249 0.316 0.030 0.544 1.576Central Banks, Gov’ts -0.037 0.152 0.368 0.615 1.187 1.262 0.264 -0.392
Large Corporations -0.024 -0.011 -0.003 0.049 0.033 0.078 0.076 0.186**Middle-Market Corp. -0.023 0.138 0.151 -0.254 -0.075 0.137 1.424** 3.059**
BanksRoyal Bank Internal 0.018 0.073* 0.159*** 0.335*** 0.348*** 0.500*** 0.323* 0.745**Global 0.108*** 0.309*** 0.440*** 0.935*** 0.842*** 1.095** 0.795 1.109Regional 0.207*** 0.489*** 0.832*** 2.079*** 2.063*** 1.585** 2.316*** 2.298*Customer-Service 0.026 0.117* 0.287** 0.605*** 0.734*** 0.720** 0.492 1.332**
Take-Profit Orders
End-UsersLeveraged Investors 0.014 0.004 -0.017 -0.075 -0.098 -0.686** -0.341 -0.933Institutional Investors -0.030 -0.031 0.076 0.219 0.165 -0.329 -0.444 -0.451Broker-Dealers 0.072 -0.058 -0.031 -0.431 -0.689 -0.432 0.038 -0.395Central Banks, Gov’ts 0.036* -0.201*** -0.231*** -0.085 -0.372** -0.010 -0.235 0.013Large Corp. 0.008 0.004 -0.029 -0.018 0.189 0.333** 0.287 0.301Middle-Market Corp. -0.134** -0.217* -0.211 -0.182 -0.167 -0.304 -0.318 -1.404
BanksRoyal Bank Internal 0.009 -0.011 -0.036 -0.072 -0.089 0.069 0.336 -0.129Global 0.053 0.179 0.221 0.614* 0.744* 0.588 -0.039 -1.044Regional -0.014 -0.324** -0.121 0.085 -0.120 -0.019 0.397 2.155**Customer-Service -0.050 -0.044 -0.070 -0.092 -0.046 -0.037 0.608 -0.487
Prob > F 0.0019 0.0001 0 0 0 0.0007 0.0368 0.0074R-squared 0.0032 0.0064 0.0085 0.0088 0.0062 0.0046 0.0026 0.0028
22
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 23/28
Table 5. Price Impact of Price-Contingent Orders: All Currencies, Log-Linear Estimates
This table summarizes the results of the regressions testing for a difference in the price impact of theorders from different client types:
t i
ii
it k t ss ε δ β ++=− ∑∑==
+ )log(*)log(*10000*)(8
1
8
1
TP ti,
SL ti, V V ,
where st is the log exchange rate expressed as foreign currency units per USD. k is the time horizon overwhich we measure price impact. We take the log of the order size and then reassign the buy-sell direction.The resulting variable is log (V i,t
x), where x ∈ (SL,TP). It is positive (negative) if the client instructs thedealer to buy the commodity (denominator) currency. * 10 percent significance; ** 5 percentsignificance; *** 1 percent significance.
5 min 30 min 1 hour 6 hours 12 hours 1 day 1 week 2 weeksStop-Loss Orders
End-UsersLeveraged Investors 0.399 1.849*** 1.909*** 4.306*** 3.184* 3.295 2.638 2.249
Institutional Investors -0.096 -0.520 -1.281 -1.138 -2.492 -0.485 0.631 2.312Broker-Dealers -0.041 0.562 0.637 1.402 1.390 0.417 2.747 7.810*Central Banks, Gov’ts 0.090 1.152 2.201 3.928 6.069 5.301 0.440 -1.468Large Corporations 0.023 0.365 0.444 -0.149 -1.672 -1.371 -2.742 -1.474Middle-Market Corp. -0.422 0.582 0.967 -0.963 0.682 -0.131 5.051 5.873
BanksRoyal Bank Internal 0.156 0.644*** 1.229*** 2.428*** 3.009*** 3.313*** 2.237* 6.094***Global 0.613*** 1.606*** 2.268*** 4.099*** 3.725*** 4.679** 3.003 5.680Regional 0.209 0.924** 2.021*** 3.896*** 4.317*** 4.599*** 4.214** 8.514**Customer-Service 0.894*** 2.185*** 3.925*** 8.686*** 8.834*** 5.554* 9.109** 8.284
Take-Profit OrdersEnd-Users
Leveraged Investors 0.111 -0.050 -0.347 -1.060 -1.257 -4.335** -3.657* -6.855*Institutional Investors -0.312 -0.439 0.051 1.170 0.367 -2.918 -3.282 -1.331Broker-Dealers 0.309 -0.451 -0.204 -0.703 -1.466 -0.481 1.236 2.032Central Banks, Gov’ts 0.125 -1.402** -1.995** -0.824 -3.812* -1.400 -3.824 2.243Large Corp. 0.155 0.025 -0.028 0.685 1.322 1.071 1.123 3.717Middle-Market Corp. -0.103 0.069 0.478 1.400 1.538 1.836 1.448 0.319
BanksRoyal Bank Internal 0.000 -0.004 -0.162 -0.494 -0.451 0.739 1.874 -1.503Global 0.311 0.643 0.638 2.379 3.375 3.156 -0.365 -3.964Regional -0.265 -0.157 -0.331 -0.051 0.367 0.635 3.168 -2.523Customer-Service 0.027 -0.659 0.017 0.789 -0.804 0.049 0.283 8.886
Prob > F 0.0218 0 0 0 0 0.0007 0.065 0.0491R-squared 0.0031 0.0071 0.0105 0.0093 0.0072 0.0047 0.0029 0.0031
23
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 24/28
Figure 1 A. Distribution of placed and executed orders over the day, USD/EURO
2 4 6 8 10 12 14 16 18 20 22 240
200
400
600
800
1000
1200
1400
1600
1800
2000
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of placed orders over the day, USD/EURO, GMT time
T/PS/L
All orders
2 4 6 8 10 12 14 16 18 20 22 240
50
100
150
200
250
300
350
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of executed orders over the day, USD/EURO, GMT time
T/PS/L
All orders
24
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 25/28
Figure 1 B. Distribution of placed and executed orders over the day, USD/GBP
2 4 6 8 10 12 14 16 18 20 22 240
200
400
600
800
1000
1200
1400
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of placed orders over the day, USD/GBP, GMT time
T/PS/L
All orders
2 4 6 8 10 12 14 16 18 20 22 240
20
40
60
80
100
120
140
160
180
200
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of executed orders over the day, USD/GBP, GMT time
T/PS/L
All orders
25
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 26/28
Figure 1 C. Distribution of placed and executed orders over the day, JPY/USD
2 4 6 8 10 12 14 16 18 20 22 240
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of placed orders over the day, JPYUSD, GMT time
T/PS/L
All orders
2 4 6 8 10 12 14 16 18 20 22 240
50
100
150
200
250
300
350
Hour, GMT
N u m
b e r o f o r d e r s
Distribution of executed orders over the day, JPYUSD, GMT time
T/PS/L
All orders
26
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 27/28
Figure 2. Intraday pattern of exchange rate volatility
2 4 6 8 10 12 14 16 18 20 22 240
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 10
-3 USD/GBP intraday volatility, GMT time
Hour, GMT
H o u r l y v o l a t i l i t y
2 4 6 8 10 12 14 16 18 20 22 240
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2 x 10-3
JPY/USD intraday volatility, GMT time
Hour, GMT
H o u r l y v o l a t i l i t y
2 4 6 8 10 12 14 16 18 20 22 240
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 10
-3 USD/EURO intraday volatility, GMT time
Hour, GMT
H o u r l y
v o l a t i l i t y
27
8/12/2019 Hedge Funds and the Origins of Private Information in Currency Markets
http://slidepdf.com/reader/full/hedge-funds-and-the-origins-of-private-information-in-currency-markets 28/28
Figure 3: Price impact of informed orders
Figure shows estimated coefficients from regressions of the following equation:
, where st is the log exchange rate expressed as
foreign currency units per USD. k is the time horizon over which we measure price impact. is thesigned dollar value of stop-loss order of customer of the type i; is the signed dollar value of take-
profit order of customer of the type i.
t i
ii
it k t V ss ε δ β ++=− ∑∑==
+TP ti,
SL ti, V **10000*)(
8
1
8
1SLti,V
TPti,V
4A: Price Impact of “Informed” Stop-Loss Orders, Linear EstimatesBased on point estimates only.
Dealers: Price Impact
0
5
10
15
20
25
5 min 1/2 hr 1 hr 6 hr 12 hr 1 day 1 wk 2 wks
B a s
i s P o i n
t s
( f o r
$ 1 0 m
i l l i o n
)
Global LocalInternal RegionalLeveraged Investors
4B: Price Impact of Leveraged Investor Stop-Loss Orders, Linear vs. Concave Estimates
Price Impact, Leveraged Investor Stop-Loss(at 6 hours)
0
5
10
15
20
25
30
1 5 10 15 20 25 30 35 40 45 50 55 60
Trade Size ($ millions )
B a s
i s P o
i n t s
Linear Concave